Customer review authoring assistant

• US 8,650,023 B2
• Filed: 03/21/2011
• Issued: 02/11/2014
• Est. Priority Date: 03/21/2011
• Status: Expired due to Fees

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Brun, Caroline
• Primary Examiner(s): Godbold, Douglas
• Assistant Examiner(s): ESTES, ERNEST LANGSTON
• Application Number:
  US13/052,774
• Publication Number:
  US 20120245924A1
• Time in Patent Office: 1,058 Days
• Field of Search: 704 1- 10
• US Class Current: 704/9
• CPC Class Codes: G06F 40/258 Heading extraction; Automatic titling; Numbering
• CPC Class Codes: G06F 40/279 Recognition of textual entities

Abstract

An authoring assistant includes a parser which automatically identifies opinion expressions in input text. The text may include an author'"'"'s review of an item, such as a product or service. A computer-implemented opinion review component generates an analysis of the text, which is based on the identified opinion expressions. The opinion review component computes an effective opinion of the text as a function of a measure of polarity associated with the identified opinion expressions. A representation generator generates a representation of the analysis for display on an associated user interface. The representation of the analysis includes a representation of the effective opinion. In the case of a review, the authoring assistant may allow the author to modify the review to reduce incoherence with a rating of the item.

Claims

1. An authoring assistant comprising:

memory which stores a polar vocabulary in which a polarity measure is associated with each of a set of adjectival terms, the polarity measure being based on ratings of reviews in a corpus of reviews from which the respective adjectival term was extracted;

a parser which identifies opinion expressions in an input text;

an opinion review component which generates an analysis of the text based on the identified opinion expressions, the opinion review component computing an effective opinion of the text as a function of a measure of polarity associated with the identified opinion expressions, which is based on the polarity measure associated with respective adjectival terms from the polar vocabulary that are in the opinion expressions, the opinion review component comprising a comparator component which compares the effective opinion with a rating associated with the text to determine whether the text and the rating are coherent;

a synonym generator for identifying synonyms of terms of identified opinion expressions;

a representation generator which generates a representation of the analysis for display on an associated user interface, the representation of the analysis including a representation of the effective opinion, the representation generator displaying at least one synonym of one of the terms in an identified opinion expression when the effective opinion and the rating associated with the text are incoherent; and

a processor which implements the opinion review component and representation generator.

2. The authoring assistant of claim 1, further comprising a warning component which outputs a warning when the effective opinion and rating are incoherent.

3. The authoring assistant of claim 1, wherein the text and rating are extracted from an author'"'"'s review of an item.

4. The authoring assistant of claim 1, further comprising a display device for displaying the representation.

5. The authoring assistant of claim 1, wherein in the representation, each of the identified opinion expressions is identified.

6. The authoring assistant of claim 1, further comprising a user interface which displays the representation and provides for an author to modify the text after the representation has been displayed.

7. The authoring assistant of claim 1, wherein the polarity measure is selected from a negative polarity and a positive polarity.

8. The authoring assistant of claim 1, wherein in the representation, the effective opinion is graphically represented.

9. The authoring assistant of claim 1, wherein at least some of the opinion expressions identifiable by the parser include a semantic relation between a first text element and a second text element, wherein at least one of the text elements is associated with a respective polarity measure.

10. An authoring assistant comprising:

memory which stores a polar vocabulary in which a polarity measure is associated with each of a set of adjectival terms, the polarity measure being based on ratings of reviews in a corpus of reviews from which the respective adjectival term was extracted;

a parser which identifies opinion expressions in text of an input author'"'"'s review of an item, the review including the author'"'"'s rating of the item on a predefined scale;

an opinion review component which generates an analysis of the text of the author'"'"'s review based on the identified opinion expressions, the opinion review component computing an effective opinion of the text as a function of a measure of polarity associated with the identified opinion expressions, which is based on the polarity measure associated with respective adjectival terms from the polar vocabulary that are in the opinion expressions, the opinion review component comprising a comparator component which compares the effective opinion with the author'"'"'s rating associated with the text to determine whether the text and the author'"'"'s rating of the item are coherent;

a representation generator which generates a representation of the analysis for display on an associated user interface, the representation of the analysis including a representation of the effective opinion;

a warning component which outputs a warning when the effective opinion and author'"'"'s rating are incoherent; and

a processor which implements the opinion review component and representation generator.

11. The authoring assistant of claim 10, further comprising a synonym generator for identifying synonyms of terms of identified opinion expressions, the representation generator displaying at least one synonym of one of the terms in an identified opinion expression when the effective opinion and the rating associated with the text are incoherent.

12. A method for assisting an author, comprising:

receiving an author'"'"'s review of an item which includes text and an associated author'"'"'s rating of the item on a predefined scale;

parsing the text of the author'"'"'s review of the item to identify opinion expressions in the input text;

with a processor, generating an analysis of the text, based on the identified opinion expressions including computing an effective opinion of the text as a function of a measure of polarity associated with each of the identified opinion expressions, the measure of polarity being is based on the polarity measure associated with respective adjectival terms from a polar vocabulary that are in the opinion expressions, the polar vocabulary associating a polarity measure with each of a set of adjectival terms, the polarity measure being based on ratings of reviews in a corpus of reviews from which the respective adjectival term was extracted;

comparing the effective opinion with the author'"'"'s rating to determine whether the text and the author'"'"'s rating are coherent; and

generating a representation of the analysis for display on a user interface, the representation of the analysis including a representation of the effective opinion.

13. The method of claim 12, further comprising displaying the representation of the user interface and providing for an author to modify the text after the representation has been displayed.

14. The method of claim 12, further comprising outputting a warning when the effective opinion and rating are incoherent.

15. The method of claim 12, further comprising providing for identifying synonyms of terms of identified opinion expressions and for displaying at least one synonym of one of the terms in an identified opinion expression.

16. The method of claim 15, further comprising displaying at least one identified synonym of one of the terms in an identified opinion expression when the effective opinion and the rating associated with the text are incoherent.

17. The method of claim 12, wherein in the representation, each of the identified opinion expressions is identified.

18. The method of claim 12, further comprising storing the polar vocabulary comprising the set of polar terms in memory, each polar term in the polar vocabulary being associated with a polarity measure, the measure of polarity associated with at least some of the identified opinion expressions being based on the polarity measure of one of the polar terms which is present in the opinion expression.

19. The method of claim 18, further comprising acquiring a set of polar adjectival terms from a corpus of reviews, each of the reviews in the corpus including text and an associated rating provided by the author of the review and incorporating the polar adjectives into the polar vocabulary.

20. The method of claim 19, wherein the acquiring of the set of polar adjectival terms comprises:

extracting textual content from reviews in the corpus of reviews;

identifying a set of frequent nouns from the textual content of the reviews,

extracting adjectival terms from the textual content of the reviews, each adjectival term being associated in the textual content with one of the frequent nouns; and

generating the set of polar adjectival terms including at least some of the extracted adjectival terms, a polarity measure being associated with each adjectival term in the set which is based on the ratings of the reviews from which the adjectival term was extracted.

21. The method of claim 18, wherein when an opinion expression is identified as a negated opinion expression, the polarity measure of the polar term is modified.

22. The method of claim 12, wherein the polarity measure associated with polar terms in the polar vocabulary is selected from a negative polarity and a positive polarity.

23. The method of claim 12, wherein in the representation, the effective opinion is graphically represented.

24. The method of claim 12, wherein the extraction of the opinion expressions includes applying parser rules for identifying a semantic relation between a first text element and a second text element, wherein at least one of the text elements is associated with a polarity measure upon which the measure of polarity of the opinion expression is based.

25. A system for performing the method of claim 12 comprising memory which stores instructions for performing the method and a processor in communication with the memory for executing the instructions.

26. A computer program product comprising a non-transitory recording medium encoding instructions, which when executed on a computer causes the computer to perform the method of claim 12.

27. A method comprising:

extracting textual content from reviews in a corpus of reviews, each of the reviews including a rating on a predefined scale by the author of the textual content;

identifying a set of frequent nouns from the textual content of the reviews;

extracting adjectival terms from the textual content of the reviews, each adjectival term being associated in the textual content with one of the frequent nouns;

generating a polar vocabulary which includes at least some of the extracted adjectival terms, a measure of polarity being associated with each adjectival term in the polar vocabulary which is based on the ratings of the reviews from which the adjectival term was extracted;

providing for parsing an input text to identify opinion expressions in the input text;

with a processor, providing for generating an analysis of the text, based on the identified opinion expressions including computing an effective opinion of the text as a function of a measure of polarity associated with each of the identified opinion expressions which is based on the measure of polarity associated with respective adjectival terms from the polar vocabulary that are in the opinion expressions; and

generating a representation of the analysis for display on a user interface, the representation of the analysis including a representation of the effective opinion.

28. The method of claim 27, wherein the generating of the polar vocabulary comprises identifying a set of positive reviews and a set of negative reviews, based on the ratings, and computing, for an identified adjectival term, a measure of its occurrence in the positive and negative sets of reviews, the polarity measure of the term being based on the measure of occurrence.